Diff command

#1
Hello everyone,

I have been reading information about the diff command in the following article:

https://journals.sagepub.com/doi/pdf/10.1177/1536867X1601600108

And practicing with some data of my own... but I have a doubt and I wonder if anyone could help me to solve it. There I go:

When I run a DID with covariates using the test command (point 4.2 within the article) and use the test option, I get different p-values than if I run a ttest in period t=0.

I could understand it if I used kernel propensity score matching (as in point 4.3 within the article) because there the ttest is estimated through linear regression, but not in the previous case (since there is no matching).

Any help?
Thank you.
 
#2
I found the solution, so I answer myself in case someone would find it helpful.

It is not true that I get different p-values. In my series there was a couple of missing values in one of the variables, so -apparently- Stata is not using some observation to perform the DID. Thus, the group that I considered to perform the initial ttest is not the same that Stata is considering to perform the ttest after DID.

Hope this could help anyone.